Voice cloning gets a bad reputation — scams, deepfakes, fraud. But the same technology has a profoundly human use: giving people with ALS and other degenerative conditions a way to keep speaking in their own voice, even after the disease takes their ability to talk.
How it works
The practice is called voice banking. While someone can still speak clearly, they record samples of their speech, and AI builds a synthesized copy of their voice. As their condition progresses and they rely on text-to-speech to communicate, the device speaks in their voice — not the flat, robotic generic voice that assistive tech offered for decades. Experts urge patients to start early, while the voice is at its strongest.
A dramatic leap in 2026
The technology has improved fast. Models that needed about 30 minutes of recorded audio in early 2025 can now produce comparable results from under a minute of speech. That matters enormously for patients whose voices are already weakening — the bar to preserve a usable voice has dropped to almost nothing.
Who’s making it accessible
Nonprofits are removing the cost barrier. In Canada, Bridging Voice partners with ElevenLabs to provide free voice clones to people with ALS; in the U.S., Team Gleason works with voice-banking companies and funds the software for anyone diagnosed with ALS. New tools keep arriving — the app “Talk To Me, Goose!” was recognized as a 2026 awardee at the Zero Project Conference for its work in this space.
The caveat
The very power that makes this moving also demands guardrails: a cloned voice is sensitive, and consent, security and control over who can use it are paramount. And while it preserves the sound of a person’s voice, communication still depends on the speed and ease of the text-to-speech interface — an area where work continues.
Why it matters
A voice is bound up with identity — it’s how loved ones recognize you, how your children remember you. Losing it to ALS is a second grief layered on top of the disease. AI voice cloning, offered free by nonprofits and now needing barely a minute of audio, lets people keep that piece of themselves. It’s the clearest answer to the question: can voice cloning be used for good?
